Threshold Effect of High-Tech Industrial Scale on Green Development—Evidence from Yangtze River Economic Belt

Abstract

Based on the panel data of 11 regions in the Yangtze River Economic Belt from 1998 to 2016, we tested and analyzed the effects of high-tech industrial expansion on green development. For these regions in the Yangtze River Economic Belt, we wanted to investigate the potential linear relationship between the scale of high-tech industry and green development or the possible threshold effect. We wanted to determine if this relationship is different in various regions of the Yangtze River Economic Belt. According to the empirical test, we found that: (1) for the entire Yangtze River Economic Belt region, the influence of high-tech industrial scale on green development doubled the threshold effect, and a marginal efficiency diminishing effect existed with the further increase in scale; (2) due to the differences among the regions, the threshold effect was different in different regions, with a double threshold effect in the lower reaches, a single threshold effect in the middle reaches, and no threshold effect in the upper reaches; and (3) regarding the high-tech industrial scale, the downstream areas were too large to weaken its promoting effect on green development. In the middle reaches, the positive impact on green development was still increasing, and the high-tech industrial scale should be further expanded. However, in the upstream areas, high-tech industrial scales did not reach the threshold value and the relationship between the high-tech industrial scale and green development was linear. Therefore, local high-tech industries should be cultivated and developed.

1. Introduction

The concept of the green economy was first proposed by Pearce, a British environmental economist, from his Blue Book on Green Economies in 1989 [1]. However, the concept of the green economy has transformed from a simple economic term into a basic consensus of human social development with the emergence of a series of problems such as climate warming, environmental degradation, and resource shortage [2]. Now, more attention is focused on environmental problems with an emphasis on green development, rather than solely aiming for an economic increase. According to the Sixth Population Census (2010), the Chinese population has reached 1.39 billion, which accounts for 19.27% of the world’s population [3]. Therefore, green development in China will greatly contribute to sustainable development worldwide. Against the background of a severe resource availability crisis and increasing environmental degradation in China, governments are committed to sustainable development and green development in order to mitigate the contradiction between economic development and the resource environment [4,5]. The high-technology industry consumes resources at a low level, which plays a crucial role in promoting upgrades of the industrial structure and improving labor productivity, and thus economic benefits [6,7]. Additionally, high-tech industries themselves are low-carbon industries based on green and sustainable development [8]. As a result, we believe that there is a close relationship between the high-tech industry and green development [9]. Lee et al. [8] defined high-tech industry as research and development (R&D)-oriented industries that possess intensive high technologies. From the Chinese perspective, high-tech industries can be divided into information technology, bioengineering technology, aerospace technology, advanced manufacturing technology, marine technology, new medicine, new materials, and new energy [10]. Compared with traditional industries, high-tech industries consume fewer material resources and have lower labor costs [11].

The research on high-tech industry and green development originated from the relationship between high-tech industry and economic growth [12]. Some studies hypothesized that the development of high-tech industry plays a crucial role in economic growth, transformation, and upgrading the industrial structure [13,14]. Some scholars provided a more detailed analysis of the internal mechanism of promoting economic growth through the high-tech industry [15]. The growth of high-tech industries has a strong multiplier effect on driving the development of related industries [16] and accelerates the rapid growth of the national economy. Some researchers described the relationship from the perspective of industrial R&D [17,18] by establishing the endogenous economic growth model based on R&D and reported that technological progress was the endogenous source of long-term economic growth. Scholars have shown that high-tech industries and economic growth are mutually driven [19,20]. The agglomeration effect of high-tech industries also has a great impact on economic growth. Knowledge spillover and local protectionism are decisive factors in high-tech industrial agglomeration in China [21], which has indirectly stunted economic growth. Shankar [22] concluded that the institutional environment is an important factor for knowledge production efficiency in high-tech industries, and this effect has significant regional differences.

Many scholars have studied the innovation performance and factor input of the high-tech industry [23,24,25]. However, with the strengthening of people’s awareness of environmental protection, researchers started to consider ecological and environmental factors in their studies. Albino [26] investigated whether the development of green products was supported by the environmental strategic approaches adopted by sustainability-driven companies, and whether there were economic sectors or geographical area specificities. Aldieri [27] analyzed the role of the knowledge diffusion process in the employment effects of sustainable development investments for large international firms. Yang [28] used the DEA (data envelopment analysis) model to determine the green development frontier surface based on the annual cross-section data from 31 regions from 2008 to 2012, and discussed how the results may allude to a superior green development pathway for China. From the perspective of green development in studying the high-tech industry, Law [29] focused on a study undertaken among a number of high-tech manufacturing firms to identify the key factors motivating the adoption and implementation of sustainability development strategies in Hong Kong. Huang [30] conducted a large sample survey of high-tech industries, including the electrical, electronics, and information industries in Taiwan, to examine organizational factor effects on GNP (Gross National Product) and the relationship between green product innovation performance and financial performance. Chesbrough [31] concluded that open innovation had utility as a paradigm for industrial innovation beyond high-tech to more traditional and mature industries, providing sustainable growth in the high-tech industry.

The above studies show that a portion of research focused on the agglomeration of high-tech industries, innovation performance, and the R&D investment in high-tech industries. Some researchers measured green development using the DEA model. Other scholars explored the relationship between high-tech industries and green development, but these studies explained the relationship in a qualitative method or from the aspect of mechanism analysis and did not use precise mathematical methods. Their research studied whether the high-tech industry had a positive effect on green development or promoted green development. From the micro perspective, technological innovation is the process of high-tech industrialization and high-tech enterprise growth, forming a certain industrial scale [32]. At the same time, technological innovation is an indispensable support for green development. It will accelerate the generation of green technology, which can realize the recycling and comprehensive utilization of resources, optimize the energy structure, and improve the utilization rate of resources [33,34]. In general, it is beneficial to promote green development. This effect or relationship is described in Figure 1. Therefore, based on previous studies and theoretical analysis, we studied the effect between the high-tech industry and green development. First, with the expansion of industry, the influence of industry will increase [35]. We considered that the expansion of the high-tech industry could strengthen the positive influence on green development. We inferred that the relationship between high-tech industry and green development was linear. With further reflection, we do not think that the relationship can maintain this linear state with the continued expansion at the industrial scale. Next, we hypothesized that there is a non-linear relationship between the scale of high-tech industry and green development and one or more thresholds may exist. Hence, we used the threshold regression model to verify the relationship in 11 provinces (cities) in the Yangtze River Economic Belt (hereinafter referred to as YREB) from 1998 to 2016.

The YREB includes a large number of energy-intensive and high-tech industries, which straddles the central and western parts of China [36]. For these regions in the YREB, we wanted to know if the relationship between the scale of high-tech industry and green development is linear or if a threshold effect exists. Additionally, we wanted to know whether the relationship would be different in various regions of the YREB.

The structure of the remaining sections is as follows: Section 2 presents the measure of green development; Section 3 provides the threshold model and variable description; Section 4 outlines the empirical analysis; and the last section includes our conclusions and recommendations.

2. Measure of Green Development

Green development originated from the concept of sustainable development. The phrase “sustainable development” reflects the prominence at the end of the 20th century and the beginning of the 21st century of the problem of acute global poverty and global environmental degradation. Although these crises are linked, problems of environment and development are often addressed independently [37]. Gradually, sustainable development has become a central concept in development studies, building on environmental, social, and political development theory and practice [38]—hence the term we now call “green development”. Green development is an abstract concept and hard to quantify. To more precisely analyze the relationship between the high-tech industry and green development, we used the Super-SBM (slack based measure) DEA model to quantify green development before analysis. The result measured by the DEA model is the efficiency of green development, which represents the degree of green development.

2.1. Super-SBM DEA Model

To measure green development, we used the input–output model to calculate the efficiency of green development. For the calculation of green development efficiency, we considered that some undesirable outputs exist according to the actual situation, such as environmental pollution with economic growth, so scholars have generally adopted the SBM-DEA model with undesirable outputs to solve these issues [39,40,41]. To compare the green development efficiency at the frontier, we adopted the undesirable output Super-SBM DEA model proposed by Tone [42]. The assumptions of the Super-SBM DEA model with undesirable outputs are as follows:

We assumed that there were n DMU composed of energy input m₁, non-energy input m₂, desirable output r₁, and undesirable output r₂, where the matrix is defined as:

$$\begin{array}{l}{\mathit{X}}^e=[x_1,\dots ,x_n]\in {\mathit{R}}^{m_1\times n}\\ {\mathit{X}}^a=[x_1,\dots ,x_n]\in {\mathit{R}}^{m_2\times n}\\ {\mathit{Y}}^d=[y_1^d,\dots ,y_n^d]\in {\mathit{R}}^{r_1\times n}\\ {\mathit{Y}}^u=[y_1^u,\dots ,y_n^u]\in {\mathit{R}}^{r_2\times n}\end{array}$$

(1)

The Super-SBM DEA model with undesirable outputs is:

$$\begin{array}{l}\min \rho =\frac{1-1/(m_1+m_2)({\displaystyle \sum _{h=1}^{m_1}{x}_h^{-e}/x_{hk}+{\displaystyle \sum _{i=1}^{m_2}{x}_i^{-a}/x_{ik}})}}{1+1/(r_1+r_2)({\displaystyle \sum _{s=1}^{r_1}{y}_s^{+d}/y_{sk}}+{\displaystyle \sum _{q=1}^{r_2}{y}_q^{-u}/y_{qk}})}\\ \begin{array}{ll}s.t.& {\displaystyle \sum _{j=1,\ne k}^n{x}_{hj}}{\lambda }_j-x_h^{-e}\le x_{h0}h=1,\dots ,m_1\\ & {\displaystyle \sum _{j=1,\ne k}^n{x}_{ij}}{\lambda }_j-x_i^{-a}\le x_{i0}i=1,\dots ,m_2\\ & {\displaystyle \sum _{j=1,\ne k}^n{y}_{sj}}{\lambda }_j+y_s^{+d}\ge y_{s0}s=1,\dots ,r_1\\ & {\displaystyle \sum _{j=1,\ne k}^n{y}_{qj}}{\lambda }_j-y_q^{-u}\le y_{q0}q=1,\dots ,r_2\\ & x_h^{-e},x_i^{-a},y_s^{+d},y_q^{-u},{\lambda }_j\ge 0j=1,\dots ,n,j\ne k\end{array}\end{array}$$

(2)

where $\rho$ is the efficiency value that seeks to increase the desirable outputs as the maximum reduction of inputs and undesirable outputs. The average ratio between the reduction of inputs and undesirable outputs and the increase of desirable outputs is used to measure the efficiency of green development. If $\rho$ < 1, then DUM₀ is invalid; if $\rho$ ≥ 1, then DUM₀ is effective. The larger the value, the higher the efficiency. There are j = 1, …, n DMU_j. $x_h^{-e},x_i^{-a},y_s^{+d},y_q^{-u}$ are slack variables of energy input, non-energy input, desirable output, and undesirable output, respectively. ${\lambda }_j$ is the intensity variable. $x_{hj},x_{ij},y_{sj},y_{qj}$ are the energy input variable, non-energy input, desirable output, and undesirable output of DMU_j, respectively. $x_{h0}$, $x_{i0}$, $y_{s0}$, $y_{q0}$ are the energy input, non-energy input, desirable output, and undesirable output of DMU₀, respectively.

2.2. Input–Output Indicators

Considering the availability and contrast of the data, we selected the input–output data of 11 provinces and cities from 1998 to 2016 to measure the efficiency of green development. There were two reasons why we chose these data. First, the output indicator chemical oxygen demand (COD) emissions only contained industrial COD emissions, and the domestic COD emissions were not recorded before 1998. The statistical caliber changed after 1998, which included domestic COD emissions and COD emissions in other wastewater sources. The other reason is that Chongqing has been a municipality directly under the Central Government since 1997, so the data of this city are available from 1998 to 2016. All of the data of the input–output indicators were obtained from the China Energy Statistics Yearbook, China Environment Statistics Yearbook, China Statistical Yearbook, and China Population and Employment Statistics Yearbook. The input–output indicators are as follows.

2.2.1. Energy Input

We chose the total social energy consumption of each region as the energy input (unit: 10,000 tons of standard coal).

2.2.2. Non-Energy Input

We mainly considered the labor input and capital input. Labor input is expressed by the total number of employees in each region. Capital input is expressed by the fiscal expenditure of each region.

2.2.3. Desirable Output

We selected the actual gross domestic product (GDP) of each region based on 1998 as the desirable output.

2.2.4. Undesirable Output

Considering the environmental pollution, we selected the emission of “three wastes” as the undesirable output: wastewater, exhaust gas, and residues. There are many kinds of pollutants in wastewater, such as chemical oxygen demand, ammonia nitrogen, lead, mercury, and cadmium. The exhaust gas includes sulfur dioxide, oxynitride, and dust. Residues include general industrial solid waste and hazardous waste. Sulfur dioxide, chemical oxygen demand, and general industrial solid waste are the main pollutants in exhaust gas, wastewater, and residues, respectively. Environmental protection departments use them as important indicators to measure environmental pollution. Some researchers selected the above undesirable outputs as indicators, such as the study by Yang [28]. Therefore, we chose the sulfur dioxide (SO₂) emissions in exhaust gas (unit: 10,000 tons), COD emissions in wastewater (unit: 10,000 tons), and the general industrial solid waste production in residues (unit: 10,000 tons) as the undesirable outputs.

2.3. Result of Green Development

Based on the above model and the related data, the green development efficiency of 11 provinces and cities in the YREB was measured by using MAXDEA 6.9 Pro software (Beijing Realworld Software Company Ltd., Beijing, China) under the assumption of variable returns to scale (VRS), as shown in

Table 1. Green development efficiency of 11 provinces and cities from 1998 to 2016.

	1998	1999	2000	2001	2002	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016
Guizhou	2.285	1.016	1.007	0.531	0.375	0.289	0.263	0.262	0.269	0.275	0.285	0.305	0.291	0.281	0.280	0.122	0.131	0.142	0.154
Sichuan	0.373	0.317	0.291	0.283	0.280	0.258	0.255	0.269	0.274	0.286	0.289	0.308	0.272	0.272	0.284	0.291	0.293	0.299	0.306
Yunnan	0.307	0.296	0.280	0.270	0.262	0.261	0.258	0.264	0.265	0.268	0.268	0.276	0.254	0.244	0.256	0.261	0.254	0.251	0.243
Chongqing	0.714	0.450	0.390	0.364	0.364	0.376	0.365	0.353	0.349	0.337	0.325	0.338	0.322	0.313	0.332	0.346	0.342	0.347	0.359
Anhui	1.009	0.779	0.506	0.413	0.364	0.347	0.345	0.348	0.338	0.329	0.313	0.306	0.280	0.272	0.285	0.277	0.273	0.267	0.263
Hubei	0.433	0.349	0.325	0.333	0.325	0.263	0.266	0.284	0.300	0.313	0.331	0.355	0.333	0.326	0.328	0.330	0.336	0.351	0.375
Hunan	1.011	1.013	0.550	0.466	0.371	0.325	0.317	0.322	0.328	0.338	0.360	0.384	0.360	0.352	0.364	0.342	0.344	0.351	0.370
Jiangxi	1.033	0.942	0.714	0.598	0.510	0.398	0.355	0.358	0.349	0.352	0.354	0.351	0.308	0.300	0.319	0.310	0.306	0.301	0.297
Jiangsu	0.370	0.373	0.365	0.366	0.377	0.382	0.379	0.384	0.427	0.437	0.446	0.477	0.434	0.468	0.537	0.573	0.652	0.781	1.043
Shanghai	0.251	0.263	0.275	0.302	0.305	0.325	0.351	0.373	0.397	0.424	0.423	0.448	0.455	0.485	0.522	0.499	0.529	0.575	1.003
Zhejiang	0.368	0.371	0.350	0.356	0.354	0.363	0.374	0.379	0.381	0.390	0.389	0.399	0.368	0.409	0.469	0.488	0.518	0.543	0.641

. In order to facilitate the analysis, this paper divided the YREB into three regions: the upper, middle, and lower reaches for the green development efficiency analysis (Figure 2). The upper reaches include Yunnan, Guizhou, Sichuan, and Chongqing; the middle reaches include Hunan, Hubei, Anhui, and Jiangxi; and the lower reaches include Zhejiang, Jiangsu, and Shanghai.

During 1998–2003, the green development efficiency of the middle and upper reaches showed a steep downward trend. The green development efficiency levels of the middle and upper reaches were 0.920 and 0.872 in 1998, respectively, and the efficiency levels were 0.296 and 0.333 in 2003, respectively. However, the efficiency of green development in the downstream areas increased slightly during this period, especially from 2000 to 2003. The efficiency of green development in the downstream areas was 0.330 in 1998 and 0.356 in 2003, and the value of green development efficiency in the downstream areas exceeded that in the upstream and middle regions. One of the reasons for the above phenomenon may be because after the Chinese accession to the World Trade Organization (WTO) in 2000, the downstream regions transferred most high-tech industries represented by the information industry based on their own advantages of labor, finance, and materials. Some studies found that technology-intensive industries in the upper and middle reaches are gradually gathering in the lower reaches, further optimizing the industrial structure of the lower reaches and promoting the green transformation of economic development in the lower reaches [43]. The second reason may be that in this stage, the development of industrial economy in the upper and middle reaches caused a certain degree of environmental pollution due to the undeveloped economy. Studies have shown that the implementation of regional coordinated development strategies, such as western development in 1999, the rise of economy in central region, and the revitalization of the old industrial base in northeast China, has led to the transfer of a large number of high-energy-consumption and high-pollution industries to the middle and upstream regions [44]. Many industries with high energy consumption and pollution have migrated to the middle and upper reaches, which undoubtedly have a negative impact on energy conservation, emissions reduction, and environmental protection in the middle and upper reaches. This further leads to a significant decline in green development efficiency in those regions.

From 2003 to 2016, the change trend of green development efficiency was not obvious in the middle and upper reaches, whereas the green development efficiency continued to increase slightly during 2003–2009 in the lower reaches. The efficiency of green development in the downstream areas increased significantly during 2010–2016: 0.419 in 2010 and 0.896 in 2016. Especially in 2015–2016, it showed a sharp cliff-type growth trend: 0.633 in 2015 and 0.896 in 2016. In short, the efficiency of green development gradually showed a pattern of regional polarization. The reason for this may be that the downstream regions continued to optimize their own advantages, constantly absorbing funds and talent from various regions, introducing new technologies, and further expanding the scale of high-tech industries to promote green economic growth. According to the Pollution Paradise Hypothesis, the more developed the economy, the higher the level of environmental regulation. The level of environmental regulation in the upper and middle reaches is lower than that of the developed areas in the lower reaches. Therefore, many high-polluting industrial agglomerations may exist in the middle and upper reaches, which has inhibited their green development to a certain degree. However, the downstream areas have optimized and further expanded their own advantages, which has greatly promoted green development in those regions.

The green development efficiency of the upper and middle reaches is generally higher than that of the lower reaches in the initial stage. However, with the implementation of the national development strategy and environmental protection policy, the efficiency of green development in the downstream areas has gradually increased, while the efficiency of green development in the middle and upstream areas has greatly reduced to a stable level. The gap in green development efficiency gradually increased among the downstream, middle, and upstream regions, showing a pattern of polarization known as the “scissors gap”, as shown in Figure 2. Notably, the industrial transfer brought about by the implementation of these strategies and policies not only prevents the narrowing of the economic development gap between the downstream and upstream regions, but may also make the upstream and middle regions bear most of the welfare losses related to negative environmental externalities, thus becoming a pollution refuge for the downstream regions. On the contrary, vigorous development of high-tech industries in the downstream areas has effectively driven the green growth of the economy. To some extent, the development of high-tech industry is of great significance in promoting the green development of the economy.

In this section, we used the SBM-DEA model to measure green development, which is represented by its efficiency. According to the measured results, we analyzed the green development degree of the upper, middle, and lower reaches in the YREB. In the next section, we establish the threshold model to check the relationship between high-tech industry and green development based on the results of the SBM-DEA.

3. Threshold Model and Variable Description

3.1. Threshold Model

The threshold regression model was first proposed by Hansen [45], and is widely used as an important method and tool to discuss the non-linear characteristics of macroeconomic phenomena. The single panel threshold regression model is as follows:

$$y_{it}={\theta }_1{x}_{it}+e_{it}(q_{it}\le \gamma )$$

(3)

$$y_{it}={\theta }_2{x}_{it}+e_{it}(q_{it}>\gamma )$$

(4)

where i represents each sample (i.e., each province or municipality directly under the central government), t is the year, y_it is the interpreted variable, x_it is the explanatory variable, q_it is the threshold variable, and e_it is the residual error. When the threshold variable q_it is less than or equal to the threshold value γ, the expression of the panel threshold regression model is Equation (3). When the threshold variable q_it is larger than or equal to the threshold value γ, the expression of the panel threshold regression model is Equation (4). Construct the indicator function I(·), which is evaluated as 1 when the conditions in parentheses are satisfied, and 0 if not. Combine the above two formulas into:

$$y_{it}={\theta }_1{x}_{it}I(q_{it}\le \gamma )+{\theta }_2{x}_{it}I(q_{it}>\gamma )+{\theta }^{\prime }{x}_{it}^{\prime }+{\mu }_i+e_{it}$$

(5)

where u_i represents the individual effect of each province and city, $x_{it}^{\prime }$ is the control variable, and the other variables are defined as described above. θ₁, θ₂, ${\theta }^{\prime }$, and γ are the parameters to be estimated. When estimating the threshold value, we used the threshold regression model proposed by Hansen [45], and used the bootstrap method to simulate the gradual distribution and critical value of F statistics according to the LM (Lagrange multiplier) test to determine whether there was a threshold effect. Then, the likelihood ratio statistic LR was used to determine whether the threshold effect was significant. When there were two or more thresholds for the threshold variables, we needed to repeat the above steps to continue to find other thresholds. We used 300 bootstraps. Here, the double threshold model is used as an example:

$$\begin{array}{ll}{y}_{it}& ={\theta }_1{x}_{it}I(q_{it}\le {\gamma }_1)+{\theta }_2{x}_{it}I({\gamma }_1<q_{it}\le {\gamma }_2)\\ & +{\theta }_3{x}_{it}I(q_{it}>{\gamma }_2)+{\theta }^{\prime }{x}_{it}^{\prime }+{\mu }_i+e_{it}\end{array}$$

(6)

Based on the premise of ensuring the availability of data, the interpreted variable y_it represents the green development of economy and is expressed by the index of green development efficiency (gdeff). The threshold variable q_it is the scale of high-tech industry, which is expressed by the index of high-tech industrial prime operating revenue (hightech). The scale of high-tech industry is the core explanatory variable. The control variables $x_{it}^{\prime }$ are social investment capital (K), human capital investment (humcap), industrial and economic structure (struct), openness (open), urban resident population density (popden), and technological innovation capacity (techinn). Therefore, the threshold regression model of the scale of high-tech industries in the YREB to the development of the green economy is as follows:

$$\begin{array}{c}gdef{f}_{it}={\theta }_{1}hightec{h}_{it}I(hightec{h}_{it}\le {\gamma }_1)+{\theta }_{2}hightec{h}_{it}I({\gamma }_1<q_{it}\le {\gamma }_2)\\ +{\theta }_{3}hightec{h}_{it}I(hightec{h}_{it}>{\gamma }_2)+{\theta }^{\prime }{x}_{it}^{\prime }+e_{it}\end{array}$$

(7)

3.2. Data Sources and Description of Variables

Based on the panel data of 11 provinces (cities) in the YREB region of China from 1998 to 2016, we conducted an empirical test. The data used in this study were from the China Statistics Yearbook on High Technology Industry, China Statistics Yearbook on Foreign Trade and Economic Cooperation, China Statistics Yearbook on Population and Employment, China Statistical Yearbook, and China Statistics Yearbook on Science and Technology. The data were processed logarithmically before being used for empirical testing. Considering the data acquisition and referring to some classic references [4,27,28], the control variables selected by the threshold effect model of high-tech industry on green development were as follows.

3.2.1. Social Capital (K)

The continuous investment of social capital can accelerate the creation of new technologies and products [46]; therefore, it plays a role in energy saving and emissions reduction and promotes the green development of the economy [47]. Considerable research has been published on the estimation of social capital stock, among which the most commonly used method is the Permanent Inventory Method (PIM) proposed by Goldsmith [48]. Estimating social capital stock by the PIM method, we mainly considered factors such as initial capital stock, determination of the depreciation rate, and investment deflator. Therefore, we selected the total investment of fixed assets as the investment index of the current period using the investment price index of each province (city) to perform the deflator for processing the investment data of fixed assets, and adjusted it to the actual value based on the price in 1998. The calculation method of the initial capital stock is as follows:

$$K_0=\frac{I_0}{g_i+\delta }$$

(8)

where $K_0$ represents the initial capital stock, $I_0$ represents the initial actual investment, $g_i$ is the geometric average growth rate of actual investment over a selected period of time, and δ is the depreciation rate. We studied the capital stock of the YREB from 1998 to 2016, taking 1998 as the base period. $g_i$ is the geometric average growth rate of the actual investment of the provinces (cities) from 1998 to 2016. For the determination of the depreciation rate, Wu [49] conducted relevant research on different depreciation rates for different provinces. We used this method as a reference to select the depreciation rates of different regions for calculation. The initial capital amount can be obtained using Equation (8). Then, the social capital stock of the provinces (cities) in the YREB from 1998 to 2016 can be calculated using Equation (9).

$$K_t=I_t+(1-\delta )K_{t-1}.$$

(9)

3.2.2. Human Capital (humcap)

The variable of human capital mainly refers to the degree of education [50]. We assumed that the higher the level of education, the stronger a person’s environmental awareness of energy conservation and emissions reduction. Here, we used the average years of education of the population over 6 years old to measure human capital (humcap), and divided the educational level of residents into six categories: illiteracy, primary school, junior high school, senior high school, junior college, undergraduate, and graduate students. Correspondingly, the average length of education was defined as 10 years, 6 years, 9 years, 12 years, 16 years, and 19 years, respectively. The formula for calculating the average length of the education of residents is as follows:

$$\begin{array}{c}humcap=primary\times 6+junior\times 9+senior\times 12\\ +college\times 16+postgra\times 19\end{array}$$

(10)

In Equation (10), primary, junior, senior, college, and postgra represent the proportion of residents with educational levels of primary school, junior high school, senior high school, junior college, and undergraduate and graduate students who are over 6 years old, respectively.

3.2.3. Industrial Structure (struct)

The resource utilization efficiency and management efficiency of different industries are different [51], and the level of technological progress and emissions are also different to varying degrees [9]. Therefore, a change in the industrial structure will have an important impact on green development. Whether the upgrade of industrial structures can further promote green development depends on the comparative relationship between the decline in environmental pollution caused by the increase of the proportion of the tertiary industry and the weakening effect of technological progress caused by the decrease of the proportion of the secondary industry. Therefore, we used the proportion of tertiary industry in the GDP (struct) to reflect the industrial structure.

3.2.4. Openness (open)

Openness factors have a dual impact—the degree of openness is conducive to the introduction of advanced production technology and management concepts [52], thus promoting the green development of the economy, and according to the “Pollution Haven” hypothesis [53], Chinese environmental regulation levels are lower than those of developed countries [54]. Along with the gradual improvement in the degree of openness, many pollution-intensive industries may flood into China, which will inhibit the green development of the economy [55]. In this paper, the actual annual proportion of Foreign Direct Investment (FDI) in the GDP represents the degree of openness. The amount of FDI was converted into RMB according to the annual exchange rate against the U.S. dollar, and the GDP was adjusted based on the GDP of the region in 1998.

3.2.5. Technological Innovation (techinn)

Theoretically speaking, increasing R&D investment or government expenditure on science and technology to enhance the ability of independent innovation or introduce advanced production technology and management concepts is conducive to the green development of the economy [33]. However, if the transformation ability from technological progress to environmental pollution control or enterprise production process is not strong, it may cause technological progress input to have a crowding-out effect on actual production [56]. Thus, it weakens the positive effect of technological progress on green development. Therefore, we used the intensity of regional R&D investment to measure the technological innovation level of the provinces (cities) in the YREB.

3.2.6. Population Density (popden)

To a large extent, the population density of a region can affect natural resources and the ecological environment [57]. The greater the population density, the higher the energy consumption, and pollution emissions increase accordingly [58], which affect the green development of the economy. Therefore, we used population density as a control variable, which is expressed by the number of permanent residents per square kilometer, reflecting the impact of population density on green development in different regions.

4. Empirical Analysis

4.1. Threshold Effect Test of YREB

The empirical test selected the data of provinces (cities) in the YREB from 1998 to 2016. Due to the long time span, we used the unit root test for each variable before threshold regression to avoid the problem of pseudo regression. In this study, the LLC (Levin–Lin–Chu) test was used to test the unit root of each variable. Considering that the selected data were annual data, we specified the lag order as Lag (1). The results of the unit root test are shown in

Table 2. Results of the unit root test.

Variables	Statistic Value	Prob.	Testing Method	Result
gdeff	−4.491 ***	0.000	LLC	Stationary
hightech	−1.921 **	0.018	LLC	Stationary
K	−10.291 ***	0.000	LLC	Stationary
humcap	−3.138 **	0.006	LLC	Stationary
struct	−6.299 ***	0.000	LLC	Stationary
open	−5.669 ***	0.000	LLC	Stationary
techinn	−6.679 ***	0.000	LLC	Stationary
popden	−2.301 **	0.010	LLC	Stationary

Note: ** and *** indicate statistical significance at the 5% and 1% levels, respectively.

.

From Table 2, all variables of order 1 were integrated at the 5% significance level, which rejects the original assumption that there is a unit root. Thus, all variables were stationary. Based on the above variables being stable, the Hausman test showed that the F statistic was 44.48 and the p-value was 0.000, which indicates that the fixed effect model was more reasonable. Next, we used the fixed effect panel threshold model to regress the panel data of 11 provinces (cities) in the YREB from 1998 to 2016 using Stata 14 (StataCorp LLC, College Station, Texas, USA). Hansen’s [45] Bootstrap method was used to obtain the p-value of the test statistics to determine whether there was a threshold effect. The results of the threshold effect test are as follows.

Table 3. Results of the test for threshold effects.

Thresholds	Test for Threshold Effect					Robust Test for Threshold Effect
	F-Value	Prob.	Critical Value			F-Value	Prob.	Critical Value
			10%	5%	1%			10%	5%	1%
Single Threshold	67.05	0.000	34.528	43.156	52.478	66.77	0.000	34.350	41.396	57.685
Double Threshold	53.46	0.003	25.063	40.333	50.635	52.92	0.000	40.744	50.783	76.063
Triple Threshold	9.56	0.407	31.402	108.207	155.510	11.00	0.230	23.659	88.246	156.177

demonstrates that the F-value of the LM test was 67.05 and the p-value was 0.000 when bootstrapping (BS) occurred 300 times. Therefore, the F statistic was significant at the 1% level, which rejects the hypothesis of no threshold effect at the 1% level and accepts that there is a single threshold. The threshold value of 8.884 was obtained from

Table 4. Threshold value estimates.

Thresholds	Test for Threshold Effect		Robust Test for Threshold Effect
	Threshold Value	95% Confidence Interval	Threshold Value	95% Confidence Interval
Single Threshold	8.884 ***	(3.730, 9.165)	8.867 ***	(8.860, 8.955)
Double Threshold	8.671 ***	(8.599, 8.855)	8.653 **	(8.610, 8.798)

Note: ** and *** indicate statistical significance at the 5% and 1% levels, respectively.

. Then, we conducted a double threshold test. Under the assumption of a single threshold, the F-value was 53.46, which was significant at the 1% statistic level. Thus, we rejected the assumption of a single threshold and accepted the assumption of double thresholds at a significance level of 1%. This means that there were two thresholds, 8.671 and 8.884. We then further considered if there would be three thresholds. According to the hypothesis of double thresholds, the F-value was 9.56 and the p-value was 0.407. The test results were not significant and could not reject the hypothesis of double thresholds—that is, accept the hypothesis of the double thresholds effect—and the threshold effect test stopped.

Because we used the macro-annual data of the Yangtze River Economic Zone and the time span of the data was quite long, we adopted the three-year rolling smoothing method to process data to improve the reliability of the estimation. For example, the data from 1998, 1999, and 2000 were averaged, then the data of 1999, 2000, and 2001 were averaged, and so on. The results of the robust test are shown in Table 3. After three years of rolling averages, the sample still showed the threshold effect, and the results of the double threshold effect were significant at the level of 5%. From Table 4, we can see that the differences in the robust test were small. Therefore, we believe that the results of the test for the threshold effect are reliable.

Based on the above test and analysis, the results of threshold regression are shown in

Table 5. Regression results.

Variable	Model 1	Model 2	Model 3
K	0.293 *** (0.030)	0.244 *** (0.034)	0.302 *** (0.032)
struct	0.508 *** (0.095)	0.642 *** (0.108)	0.522 *** (0.103)
humcap	0.599 *** (0.200)	0.386 (0.234)	0.953 *** (0.271)
open	−0.030 * (0.017)	−0.014 (0.019)	−0.047 *** (0.018)
techinn	−0.250 *** (0.028)	−0.235 *** (0.033)	−0.282 *** (0.032)
popden	−0.045 ** (0.016)	−0.031 * (0.018)	−0.061 *** (0.017)
hightech_1	0.078 *** (0.024)	−0.017 (0.028)	0.065 ** (0.025)
hightech_2	0.117 *** (0.023)		0.101 *** (0.024)
hightech_3	0.059 ** (0.024)		0.044 * (0.025)
constant	5.552 *** (0.395)	5.591 *** (0.442)	6.195 *** (0.434)
observation	220	220	198
R-squared	0.949	0.926	0.954

Note: Models 1, 2, and 3 are the panel threshold model with the fixed effect, the panel linear model, and the robust model, respectively. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

.

To better compare the results of threshold regression model 1, we used linear regression to obtain the results of model 2 from the panel data. From the results of model 1, the promotion effect of the high-tech industry scale on economic development was a non-linear relationship, and there were two threshold values. Specifically, the two threshold values of 8.671 and 8.884 divided the model into three stages. In the first stage, when hightech was lower than 8.671, its coefficient was 0.078; in the second stage, when hightech was larger than 8.671 and smaller than 8.884, its coefficient was 0.117; whereas in the third stage, when hightech was larger than 8.884, its coefficient was 0.059. In general, the coefficients of the three stages were all positive, but compared with the second stage, the scale of high-tech industry in the first stage was limited, which contributed less to green development and had a weaker influence than in the second stage. As the industrial scale reached a certain level, the impact on green development gradually emerged in the second stage. However, as the industrial scale continued to expand and exceed the second threshold, its role in green development slowed down. In other words, the higher scale of high-tech industry is certainly better, and its scale needs to be combined with actual demand. The scale of the high-tech industry not only has a double threshold effect, but also a diminishing marginal efficiency effect on green development. From the results of model 2, we can see that the high-tech industry scale (hightech), as a core explanatory variable and a threshold variable, was negative at the 1% significance level. This suggests that it has a significant inhibitory effect on green development with continuous expansion of the high-tech industrial scale. Thus, simply setting the regression model as a panel linear model will lead to unpredictable deviation, and even produce wrong conclusions. Therefore, we believe that there is a non-linear relationship between the scale of high-tech industries and green development.

4.2. Threshold Effect in Upper, Middle, and Lower Reaches

From the above results, we found that there was an issue with the two thresholds of the high-tech industry scale (8.671 and 8.884). For some undeveloped provinces, such as Anhui and Jiangxi, the scale of the high-tech industry did not reach the first threshold value from 1998 to 2016. The possible reason is that the large scale of the high-tech industry in some developed areas such as Jiangsu powerfully raised the threshold value. The 11 provinces (cities) in the YREB were divided into the upper, middle, and lower reaches. The upper reaches include Chongqing, Sichuan, Yunnan, and Guizhou; the middle reaches include Anhui, Hubei, Hunan, and Jiangxi; and the lower reaches include Zhejiang, Shanghai, and Jiangsu. Next, we ran threshold regression to observe whether the results of the regression had a double threshold effect in different regions.

We found that the threshold values of the upper, middle, and lower reaches of the YREB were different from the previous results of threshold regression. The results in

Table 6. The results of the test for threshold effects in different regions.

Thresholds	Upper Reaches					Middle Reaches					Lower Reaches
	F-Value	Prob.	Critical Value			F-Value	Prob.	Critical Value			F-Value	Prob.	Critical Value
			10%	5%	1%			10%	5%	1%			10%	5%	1%
Single Threshold	12.51	0.217	16.284	18.847	21.569	43.11	0.000	21.618	27.773	38.377	29.62	0.000	10.856	11.636	12.314
Double Threshold	6.13	0.440	12.349	15.576	18.201	12.52	0.248	19.499	26.497	45.539	12.55	0.027	8.870	9.191	16.546
Triple Threshold	5.76	0.537	11.675	14.589	16.684	8.00	0.355	16.153	20.781	30.736	19.51	0.167	22.502	23.239	35.976

;

Table 7. The threshold estimators and confidence intervals.

Thresholds	Upper Reaches		Middle Reaches		Lower Reaches
	Threshold Value	95% Confidence Interval	Threshold Value	95% Confidence Interval	Threshold Value	95% Confidence Interval
Single Threshold	\	\	4.683 ***	(4.619, 4.805)	5.872 ***	(5.568, 5.966)
Double Threshold	\	\	\	\	7.851 **	(7.543, 7.863)

Note: ** and *** indicate statistical significance at the 5% and 1% levels, respectively.

show that there was still a double threshold effect in the downstream region. The threshold values were 5.872 and 7.852. There was only one threshold value in the middle reaches: 4.683. However, the threshold effect was not significant in the upstream region, and there was no threshold effect. The regression results for the upper, middle, and lower reaches are shown in

Table 8. Results of the threshold model regression.

Variables	Upper Reaches	Middle Reaches	Lower Reaches
ln K	0.414 *** (0.044)	0.173 ** (0.079)	−0.156 *** (0.030)
ln struct	0.740 *** (0.176)	−0.382 ** (0.178)	1.989 *** (0.158)
ln humcap	0.134 (0.251)	0.757 *** (0.284)	0.615 ** (0.275)
ln open	−0.011 (0.022)	−0.047 (0.034)	−0.091 ** (0.041)
ln techinn	−0.453 *** (0.051)	−0.282 *** (0.062)	0.292 *** (0.060)
ln popden	−0.049 *** (0.017)	−0.029 (0.033)	−0.118 *** (0.032)
ln high-tech_1	0.066 ** (0.032)	0.074 ** (0.030)	0.085 *** (0.028)
ln hightech_2		0.183 *** (0.061)	0.057 * (0.029)
ln hightech_3			−0.006 (0.028)
constant	5.687 *** (0.635)	1.475 * (0.862)	8.245 *** (0.709)
observations	80	80	60
Hausman test	F = 59.75 P = 0.000	F = 43.79 P = 0.000	F = 27.78 P = 0.000

Note: The results in the upper, middle, and lower reaches were used in the panel linear model, single threshold model, and double threshold model, respectively. The Hausman test showed that we should use the fixed effect model. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. To eliminate heteroscedasticity, we take the logarithm of the data.

.

Based on the test results in Table 8, in the upstream regions, the coefficients of openness (open), technological innovation (techinn), and population density (popden) variables were positive. The three variables had a certain negative effect on green development. However, open was not significant. This shows that the negative effect of the Pollution Paradise Hypothesis was obvious. The coefficients of social capital (K), industrial structure (struct), and human capital (humcap) were positive. Social capital (K) and industrial structure (struct) were significant at the level of 1%, while human capital (humcap) was not significant. The coefficient of the high-tech industry scale (high-tech) was also positive, which indicates that high-tech industry promoted green development, but its effect can be further strengthened in less developed areas. The high-tech industry may have developed late and on a small scale in less developed areas, being too weak to drive green development. However, with the continuous expansion of the high-tech industrial scale, there may be a more significant impact in the future.

In the middle reaches, when the scale of high-tech industry was less than 4.683 in the first stage, the expansion of the high-tech industrial scale had a positive effect on green development and was significant at the 1% level. In the second stage, when the scale of high-tech industry exceeded 4.683, the expansion of the high-tech industrial scale had a more obvious positive effect on green development. The coefficient in the first stage was 0.074, and the coefficient in the second stage was 0.183. This shows that the expansion of high-tech industries in the middle reaches promoted green development in general, but the promotion was weak in the initial stage when the scale was small and below the threshold value. When the high-tech industrial scale exceeded the threshold, it had a significant positive impact on green development. The coefficients of the industrial structure variable (struct) were significantly negative at the 5% level and the coefficients of population density (popden) were not significant; the symbolic properties of the coefficients of other control variables were the same as those variables in the upper reaches. This suggests that the reduction in energy consumption and pollution caused by the increased proportion of tertiary industry in the middle reaches was lower than the weakening of technological progress caused by the decreased proportion of secondary industry, which is not conducive to green development.

The results in the downstream areas showed that the scale of high-tech industry (hightech) had a double threshold effect on green development. In the first stage, when the scale of high-tech industry was less than 5.872, its coefficient was 0.085 and significant at the 1% level, which indicates that high-tech industrial scale expansion has a certain role in promoting green development. In the second stage, when the scale of the high-tech industry was greater than 5.872 and less than 7.851, the coefficient was 0.057 and significant at the level of 10%, which indicates that the promotion effect of the further expansion of the high-tech industrial scale on green development was less than that in the first stage. In the third stage, when the scale of high-tech industry was greater than 7.851, the coefficient was −0.006 and not significant, which shows that the expansion of the industrial scale played a restraining role in green development in this stage. Through comparison, we found that the coefficients of high-tech industrial scale in the first and second stages were both positive, and the coefficients in the first stage were larger than that in the second stage, whereas it was negative in the third stage. This shows that the scale of high-tech industry in the downstream areas had a strong positive impact on green development when it was lower than the first threshold value. As the scale continued to expand and reached the value between the two threshold values, the promotion effect on green development declined significantly. However, if the industrial scale was greater than the second threshold value, green development could not be driven effectively by high-tech industry.

Comparing the test results of the upper, middle, and lower reaches in the YREB, we found that the coefficients of population density (popden) were all negative, and the results were not significant only in the middle reaches. The coefficients of industrial structure (struct) were negative in the middle reaches and positive in the upper and lower reaches. Notably, the coefficients of social capital (K) were significantly positive in the upper and middle reaches, which had different promotion effects on green development. However, K was negative in the developed areas of the lower reaches. This indicates that social capital has a negative effect on green development, which may be caused by the excessive capital accumulation in developed areas. This will increase the leverage and loan quota with the investment of capital, then lead to economic bubbles, which will negatively affect green development. For human capital (humcap), openness (open), and technological innovation (techinn), the coefficients of humcap were all positive in the upper, middle, and lower reaches, but not significant only in the upper reaches. In contrast, the coefficients of open in the upper, middle, and lower reaches were all negative, but only significant at the 5% level in the lower reaches. The results further verified the existence of the Pollution Paradise Hypothesis in the less developed areas. The coefficients of techinn were significant at the 1% level in the three regions, but positive only in the downstream regions. This shows that the transformation rate of scientific and technological innovations is high in developed areas, which has a positive effect on green development. However, the transformation level of scientific and technological innovation is low in less developed areas, and further improvement is required to promote green development.

5. Conclusions and Recommendations

Using the scale of high-tech industry as the threshold variable, we studied the non-linear impact of the scale of high-tech industry on green development in 11 provinces and municipalities in the Yangtze River Economic Belt (YREB) region of China from 1998 to 2016 by using the panel threshold model with a fixed effect and drew the following conclusions:

(1)

In general, the expansion of the high-tech industry scale in the YREB promoted green development and the promoting effect was different in the regions of the upper, middle, and lower reaches. We verified that the scale of high-tech industry had a double-threshold effect on the green development efficiency. The threshold values were 8.884 and 8.671. We compared this with the previous linear model (model 2) in Table 5, which shows that human capital (humcap), openness (open), and the explanatory variable high-tech industry scale were not significant. From model 1, openness (open) was significant at the 10% level; population density (popden) and high-tech industry scale in the third stage were significant at the 5% level. The rest of the variables were significant at the 1% level in model 1. The results of the robust test in model 3 were similar to the results in model 1. The threshold model (model 1) was more appropriate to analyze the relationship between high-tech industry and green development. According to the results of the threshold model, we conclude that excessive industrial scale limits the promotion effect of green development, and the greatest promotion impact is between the two threshold values (8.671 and 8.884).

(2)

In the upper reaches, there was no threshold effect as shown in Table 7, and the relationship between the high-tech industry scale and green development was linear after the regression test (the results of the non-linear test were not significant). Therefore, we used the fixed effect linear model to analyze the relationship between high-tech industry and green development. From the results in Table 8, the coefficient of the high-tech industry scale was positive and significant at the 5% level. Thus, we conclude that high-tech industry in remote and developing upstream areas has a positive effect on green development. The coefficients of social capital (K), industrial structure (struct), and human capital (humcap) were positive, and the coefficients of openness (open), technological innovation (techinn), and population density (popden) were negative. However, human capital (humcap) and openness (open) were not significant. The other control variables were significant at the 1% level. This indicates that social capital accumulation and industrial structure significantly positively impact green development. However, enhanced technological innovation and increased population density have an inhibitory effect on green development in upstream regions.

(3)

In the middle reaches, the expansion of high-tech industries had a single threshold effect on green development (Table 7), where the threshold value was 4.6830. According to the result in Table 8, high-tech industrial expansion had a positive impact in two stages (less than or greater than the threshold value). In the second stage, when the scale was greater than the threshold value, the expansion of the high-tech industry scale played a more significant role in promoting green development (the coefficient in the first stage was 0.074 and significant at the 5% level; the coefficient in the second stage was 0.183 and significant at the 1% level). The coefficients of social capital (K) and human capital (humcap) were positive, and the coefficients of industrial structure (struct), openness (open), technological innovation (techinn), and population density (popden) were negative. Among these control variables, openness (open) and population density (popden) were not significant. The rest of the variables were significant at the 5% or 1% level. This implies that social capital accumulation and improvement in the education level of residents can promote green development, but industrial structure improvement and enhanced technological innovation can be obstructions for green development in the midstream regions.

(4)

In the lower reaches, from Table 7, the expansion of the high-tech industry had a double threshold effect on green development and the threshold values were 5.872 and 7.851. In Table 8, the coefficients of the first stage and second stage were positive, so we conclude that the expansion of high-tech industry promoted green development. The coefficient of the first stage (high-tech_1) was greater than the coefficient of the second stage (high-tech_2). The decline of the high-tech coefficient indicated that the promotion of green development decreased with the expansion of the industrial scale. At the end of the time period, influence coefficient became obscure or even counterproductive, as in the third stage, the coefficient was negative and not significant. This means that high-tech industry had a weak or even negative impact on green development with further expansion of the industrial scale in the third stage. These control variables were significant at the 1% or 5% level. Among these control variables, the coefficients of social capital (K), openness (open), and population density (popden) were negative, and the coefficients of industrial structure (struct), human capital (humcap), and technological innovation (techinn) were positive. This suggests that upgrading the industrial structure, improving of the education level of residents, and enhancing technological innovation are beneficial to green development. However, social capital accumulation, strengthening openness, and increased population density negatively impacted green development in the downstream regions.

Based on the empirical test and the above conclusions, the following recommendations can be made:

(1)

The YREB is one of four high-tech industrial clusters in China. Local governments should improve the high-tech industrial innovation system in these areas to effectively combine scientific research institutions, enterprises, and other institutions to drive high-tech industrial innovation. Taking technological innovation as the key starting point to develop low-carbon industries and avoiding using high technology to develop high-carbon and high-emissions industries could not only alleviate the pressure on resources and the environment, but also promote the sustainable development of the economy, which is the essence of green development.

(2)

For the upper reaches of the YREB, the high-tech industry is still in the early stages of development. According to further analysis, we found the following three implications. First, its economic foundation is weak and the people are poor, which restrict the development of local high-tech industry to a certain extent. Second, most of the cities in the upstream region have a good natural environment and rich ecological resources. They mainly focus on the tourism industry and have ignored the importance of the high-tech industry. However, sole dependence on tourism cannot support the economic demand of the whole region. Third, due to the low education level of the residents and the lack of high-tech talent, the high-tech industry is developing slowly in this area. Therefore, in the upstream regions, local governments should accelerate the pace of precise poverty alleviation and improve people’s living standards. In order to achieve sustainable economic development, they should increase the attractiveness of talent (especially high-tech innovative talent) to expand the scale of high-tech industries and strengthen the positive impact of high-tech industries in green development.

(3)

The high-tech industry has already developed on a certain scale in the middle reaches of the region. Its promotion effect on green economic development has achieved initial results and is increasing. The current bottleneck is mainly due to the conversion rate of high-tech achievements. The middle reaches have a certain economic base and a considerable reserve of high-tech talent compared with the upstream regions. However, few scientific and technological innovations are actually applied to actual products. This low transformation rate of scientific and technological innovation ultimately limits the role of high-tech industries in promoting green development. Therefore, to realize the effective connection between high-tech products and the market and further accelerate the green development of the economy, local governments should establish a third-party platform as soon as possible and improve the mechanism to encourage the transformation of high-tech achievements.

(4)

In the downstream region, the scale of the high-tech industry has reached a relatively large scale, and could even be considered saturated. This has played an evident role in promoting economic green development. These regions have already obtained a high level of economy and a high-tech talent-intensive industry. If investment continues to increase to expand the scale, there will be a marginal declining effect or even play a counterproductive role, resulting in the waste of labor, materials, and financial resources, which violates the original intention of green development. Therefore, the key to promoting green development in developed areas is to control the scale of high-tech industry and to balance the allocation of social resources.

For developed areas, it is necessary to consider some future problems with blind development and the further expansion of high-tech industries, especially in artificial intelligence and robotics. At present, the world’s largest population is still in China, especially in developed areas, where the population density is extremely high and the proportion of high-tech talent in employment is low. With the large-scale marketization of high-tech products, the unemployment rate of low-tech employees may rise, which is another issue that should be given attention in the future.